llama.cpp

presets : fix pooling param for embedding models

#16455

Merged